HEAD at 7b1d959bc4e1cae64abdeff2c5baeb9c38c3813b = 384918

gn args:
is_component_build = true
use_goma = true
target_os = "chromeos"
enable_nacl = false

ninja -C out/Default -j 1000 -l 20 ash_unittests

out/Default/ash_unittests --gtest_filter="ShelfLayoutManagerTest.ShelfLayoutInUnifiedDesktop"

[ RUN      ] ShelfLayoutManagerTest.ShelfLayoutInUnifiedDesktop
Xlib:  extension "RANDR" missing on display ":99".
[18546:18546:0404/092655:408285248160:INFO:lock_state_controller.cc(95)] Constructing LockStateController instance 0x3418f7fde360
Xlib:  extension "RANDR" missing on display ":99".
Xlib:  extension "RANDR" missing on display ":99".
../../ash/shelf/shelf_layout_manager_unittest.cc:2366: Failure
Value of: status_area_widget->GetWindowBoundsInScreen().ToString()
  Actual: "362,453 138x47"
Expected: "353,453 147x47"
[18546:18546:0404/092655:408285604933:INFO:lock_state_controller.cc(101)] Destroying LockStateController instance 0x3418f7fde360
[  FAILED  ] ShelfLayoutManagerTest.ShelfLayoutInUnifiedDesktop (459 ms)
[----------] 1 test from ShelfLayoutManagerTest (460 ms total)

The test looks a little fragile:

  // Shelf should be in the first display's area.
  // TODO: make this test more robust against changes in font, font size.
  EXPECT_EQ("353,453 147x47",
            status_area_widget->GetWindowBoundsInScreen().ToString());

It passes on the bots. I'm not sure why. I wonder if the assertion could be changed to be more liberal, like checking that the status area widget's right edge is < 500, or some such.